C&A Tutorial

Institut Tik - Eth Zuerich

Description

Peer-to-Peer GI Arbeitsgespräch, Darmstadt, September 24, 2003Distributed Pricingin P2P Networks1David Hausheer1Swiss Federal Institute of Technology, ETH ZürichComputer Engineering and Networks Laboratory TIK, Zürich, SwitzerlandE-Mail: [e-mail] Engineering and Networks Laboratory,Swiss Federal Institute of Technology, ETH Zürich© 2003 D. Hausheer, ETHZ-TIK ETH Zürich††††OutlineMotivation for Pricing in P2PBasic P2P Pricing Protocol• Price Determination• Price Valuation• Price DisseminationPricing Design• Pricing APIsConclusions© 2003 D. Hausheer, ETHZ-TIK ETH Zürich††††Motivation for Pricing in P2PObservations in current P2P file sharing systems• There is a lot of “crap” content being offered by users• Peers are not peers (heterogeneous infrastructure)Accounting “counts” all bits the same• There is no differentiation between good bits and bad bits• It is scenario-specific (how many files are 10 min connection worth?)• Users have no control (I want to provide my files for free, but I can’t)Pricing enables differentiation, generalization and control• I give you 10 Britney Spears songs for 1 Madonna song• I can’t offer files, but I have lots of computing power• If my PC is busy, I want you to pay the doubleBut this is user-specific!• That’s why we intend the users to determine the price© 2003 D. Hausheer, ETHZ-TIK ETH Zürich†Price RequestPrice NegotiationPrice AcceptanceBasic P2P Pricing ...

Informations

• Publié par: Erzo
• Nombre de lectures: 17
• Langue: English

Extrait

©2 003 D. Hausheer, E

Peer-to-Peer GI Arbeitsgespräch, Darmstadt, September 24, 2003

Distributed Pricing in P2P Networks

David Hausheer 1 1 Swiss Federal Institute of Technology, ETH Zürich Computer Engineering and Networks Laboratory TIK, Zürich, Switzerland E-Mail: [e-mail]

HT-ZITK

Computer Engineering and Networks Laboratory, Swiss Federal Institute of Technology, ETH Zürich

ETH Zürich

©2 00

†

†

†

†

3 D .H

Outline

Motivation for Pricing in P2P

Basic P2P Pricing Protocol • Price Determination • Price Valuation • Price Dissemination

Pricing Design • Pricing APIs

Conclusions

uahseer ,TEHZ-ITK ETH Zürich

©2 00

Motivation for Pricing in P2P

† Observations in current P2P file sharing systems There is a lot of “crap” content being offered by users • • Peers are not peers (heterogeneous infrastructure) † Accounting “counts” all bits the s   ame • There is no differentiation between good bits and bad bits • It is scenario-specific (how many files are 10 min connection worth?) Users have no control (I want to provide my files for free, but I can’t) • † Pricing enables differentiation, generalization and control • I give you 10 Britney Spears songs for 1 Madonna song • I can’t offer files, but I have lots of computing power • If my PC is busy, I want you to pay the double † But this is user-specific! • That’s why we intend the users to determine the price

3 D. Hausheer, ETHZ-ITK TE HZürich

©2 00

Basic P2P Pricing Protocol

† Pricing comprises of three important mechanisms: • Price determination => How a provider determines a price • Price valuation => How a customer valuates a price • Price dissemination => How prices are communicated, negotiated and accepted between providers and customers

3 D .aHsuehre ,TEZH-ITK

Custome Peer

Price Dissemination

ETH 

Provider Peer

Zürich

©2 00

† † †

† †

3 D .H

Price Determination

Pricing is flexible but offers a set of built-in strategies to determine a price Additional strategies can be created enabling any business model Built-in mechanisms: • Pricing based on price quotes (‘market-based pricing’) – Determines a price based on price quotes from other providers • Pricing based on current demand – Calculates the price for a service based on the current load • Pricing based on user ratings – Uses rating information from other users to determine the price • Fixed pricing configured by rules – Fixed or range-limited set of price coefficients to be used A user can enable, disable and combine individual pricing strategies, or create additional strategies, configured by local rules Final prices might be negotiated between customer and provider • Support of complex auction models

ausheer, ETHZ-ITK TEH Zürich

©2 00

Price Valuation

† Pricing should reveal the value or utility of a good • A rational user buys the good, if a price is lower than expected utility • Rating is needed to know what can be expected • Rating is “pricing” done by the consumer † Pricing (and rating) need to be done in a distributed way • Only a user can valuate, e.g., the utility of a service • A user can change anything locally in his code anyway • Rules can limit the prices, but enforcement is needed (exclusion?)

 3D. Hausheer ,TEZH-TIK TEH Zürich

©2 00

Price Valuation (cont d) ’

† The underlying mechanisms are similar to price determination • A customer might actually use the same strategies and information to valuate the price for an offered service • A provider can use information from other users (he knows the good) • A customer has to rely on other information (unless he fully trusts the provider) † Ratings need to be aggregated appropriately • Only scalable if done in a distributed way (=> Eigenrep paper) • Ratings need to be weighted according to user-interests, trust-levels… † What is the incentive for a user to provide ratings? • Users might get paid for ratings

 3D. Hausheer ,TEHZ-TIK TEH Zürich

©2 00

Price Dissemination

† How prices get from providers to customers • Customer asks provider for price (pull method) • Provider regularly advertises its prices (push method) • Customers might subscribe for price advertisements † More scalable approaches: • Prices are stored in a distributed hash table • Price messages are cached at intermediate peers – How long should a price remain valid? – What about self-interested peers? • Requests are routed according to price information – Two processes at once (search & valuation) † Negotiation enables further pricing approaches • Auction models might be supported

 3D. Hausheer, ETHZ-TIK TEH Zürich

Rules Mgmt

SLA Service Descr. Acc

1

Accounting

Pricing API:

‰ Support of 1. Fixed pricing by rules 2. Pricing based on demand 3. Pricing based on negotiation 4. Pricing based on price quotes (‘market prices’) 5. Pricing based on user ratings

Other Peers

‰ Methods for pulling, posting and calculate a tariff, service rating, and approval

4 API (1) Pricing API (3)

2

5

Service Negotiation 3 Price negotiation

hcirüZ HTE KIT-30D  ©02ETHZer, ushe. Ha

Pricing Design

©2 00

Pricing API (1)

† Module internal methods, provided to pricing module instances on other peers • public void getServiceTariff (ServiceId id, boolean getUpdates, PricingListener pl); • public void postServiceTariff (PeerId pid, ServiceId sid, ServiceTariff t); • public void getServiceRating (ServiceId, boolean getUpdates, PricingListener pl); • public void postServiceRating (PeerId pid, ServiceId sid, ServiceRating r); † Default is pull method † getUpdates=true adds a peer on a list of recipients (to trigger the more efficient push method )

 3D .Hausheer ,ETHZ-TIK ET HZürich

©2 00

Pricing API (2)

† Methods provided to the service negotiation module • public void requestTariff (SLA proposedSLA, PricingListener pl); • public void approveServiceTariff (ServiceTariff t, PricingListener pl); † requestTariff calculates and returns a tariff for a proposed SLA. Service Negotiation stores the agreed tariff in the SLA, where Charging can look it up. † approveServiceTariff valuates a proposed service tariff in the negotiation process. Pricing approves ServiceTariff, offers new ServiceTariff or stops negotiation.

3 D. Hausheer ,TEZH-ITK ETH Zürich